DMS is a proven technology to extend the ultimate reach of high bit rate (e.g. 10 Gb/s) optical fiber communication systems by balancing the optical fiber's dispersion and nonlinearity. Dense WDM-DMS systems, however, tend to suffer one significant nonlinear penalty, i.e., a jitter in pulse arrival times resulting from the collisions between solitons of different channels.
When the dispersion compensation of a transmission span is accomplished with only dispersion-compensating fiber DCF, a pair of colliding pulses (from different channels) dart back and forth with respect to each other in retarded time, in a sawtooth motion typically several hundreds of picoseconds in amplitude (the dispersion of the span times the channel wavelength spacing). The net motion produced by each period of the sawtooth, governed by the path-average dispersion, is, however, quite small. Thus, before the colliding pulses finally separate, the pulses tend to interact with each other in a series of “mini-collisions”, which in total comprise an overall collision that tends to be many thousands of kilometers long.
Since the repeated partial collisions that occur in the beginning stages of such an overall collision produce a steep staircase of frequency shift (from cross-phase modulation, or XPM), the net time shifts from such collisions are large (typically ˜5 ps each when adjacent channels are involved). It is these large per-collision time shifts, compounded by the poor statistics of just a handful of collisions, that tend to create eye-closing timing jitter for distances in excess of about 5000 km.
Prior art attempts to reduce the XPM penalty include schemes for increasing the rate at which colliding pulses pass through each other. For example, one scheme attempts to introduce a one bit-period delay between adjacent channels in each 100 km span with a Mach-Zehnder fiber interferometer, or with a succession of fiber Bragg-gratings. Such attempts were able to claim a 2 dB reduction in the power penalty from XPM in relatively short (≦500 km) non-return-to-zero (NRZ) transmissions. The significant benefit of such attempts was strongly focused on the immediate neighborhood of a one-bit-period delay between two adjacent WDM channels.
Another scheme considered the use of periodic group delay (PGD) devices for reduction of XPM-induced penalties, where, for the optimal dispersion maps, there was almost no improvement from numerical simulations. An experiment performed with a different set of parameters showed an improvement of only ˜0.85 dB.
The schemes discussed above were specifically used for NRZ transmission systems, without focusing on DMS systems that are intrinsically return-to-zero (RZ). The benefit of schemes using NRZ transmission stems from reduction of amplitude noise resulting from constructive interference between pulses from repeated mini-collisions. The benefit in a scheme using dispersion-managed solitons, however, stems from the reduction of XPM-induced frequency shifts and of the consequent timing jitter. The different nature of the penalties also results in the fact that whereas the XPM penalty in prior art schemes appears only with a non-optimized dispersion map, in a DMS scheme the XPM penalty tends to be inherent even with an optimized dispersion map for DMS.
Other prior art schemes for specifically reducing the collision-induced timing jitter in dense WDM-DMS systems includes frequency-guiding filters and jitter tolerant receivers, but these techniques are complex and not yet practical to implement in realistic systems. Accordingly, a need therefore exists for a method and apparatus to effectively reduce such collision-induced timing jitter and increase the transmission rate over optical WDM-DMS systems.